Mistral AI API Pricing 2026: 11 Models From $0.10/$0.10
Complete pricing for every Mistral model — from the $0.10/$0.10 Ministral 3B edge model to the $2/$5 Magistral Medium reasoning engine. The only major AI provider with EU data residency.
TL;DR
- Cheapest: Ministral 3 3B — $0.10/$0.10 per 1M tokens (symmetric pricing, 128K context)
- Best value: Mistral Large 3 — $0.50/$1.50 per 1M tokens (open-weight flagship, 262K context)
- For code: Codestral — $0.30/$0.90 per 1M tokens (256K context, fill-in-middle)
- For reasoning: Magistral Medium — $2.00/$5.00 per 1M tokens (256K context)
- EU advantage: Regional inference with EU data processing (+10% surcharge) for GDPR compliance
All Mistral API Pricing (Aug 2026)
Mistral offers 11 models across 4 tiers. All prices are per 1M tokens. Batch API pricing is approximately 50% off standard rates.
| Model | Input $/M | Output $/M | Context | Tier | License |
|---|---|---|---|---|---|
| Ministral 3 3B | $0.10 | $0.10 | 128K | Budget | Apache 2.0 |
| Devstral Small 2 | $0.10 | $0.30 | 128K | Budget | Apache 2.0 |
| Ministral 3 8B | $0.15 | $0.15 | 128K | Budget | Apache 2.0 |
| Mistral Small 4 | $0.15 | $0.60 | 128K | Budget | Apache 2.0 |
| Ministral 3 14B | $0.20 | $0.20 | 128K | Budget | Apache 2.0 |
| Codestral | $0.30 | $0.90 | 256K | Mid | Mistral |
| Devstral 2 | $0.40 | $2.00 | 128K | Budget | Apache 2.0 |
| Magistral Small | $0.50 | $1.50 | 256K | Budget | Mistral |
| Mistral Large 3 | $0.50 | $1.50 | 262K | Budget | Apache 2.0 |
| Mistral Medium 3.5 | $1.50 | $7.50 | 128K | Mid | Mistral |
| Magistral Medium | $2.00 | $5.00 | 256K | Mid | Mistral |
EU Data Residency: Mistral is the only major AI provider offering regional inference with EU data processing. Add 10% to standard pricing for GDPR-compliant EU data residency. Critical for healthcare, finance, and government workloads.
Mistral vs OpenAI vs Google: Price Comparison
How Mistral's flagship models compare to equivalent models from other providers:
| Use Case | Mistral | OpenAI | Cheapest | |
|---|---|---|---|---|
| Flagship (complex tasks) | Large 3: $0.50/$1.50 | GPT-5.4: $2.50/$15 | Gemini 3.1 Pro: $2/$12 | Mistral (80% cheaper) |
| Budget (high-volume) | Small 4: $0.15/$0.60 | GPT-5.4 nano: $0.20/$1.25 | Gemini 3 Flash: $0.50/$3 | Mistral (75% cheaper) |
| Edge/Embedded | Ministral 3B: $0.10/$0.10 | GPT-5 nano: $0.05/$0.40 | Gemini 2.5 Flash-Lite: $0.10/$0.40 | Mistral (output) |
| Code generation | Codestral: $0.30/$0.90 | GPT-5.3 Codex: $1.75/$14 | — | Mistral (83% cheaper) |
Mistral Model Family: Which One Should You Pick?
Mistral's 11 models cover every use case from embedded edge devices to complex reasoning. Here's a decision guide:
🟢 Ministral 3 3B — $0.10/$0.10
Ultra-lightweight 3B parameter model for edge devices, embedded systems, and high-volume classification. Symmetric pricing (1:1 input:output ratio) makes it the cheapest option for output-heavy tasks. Apache 2.0 licensed — self-host for zero API costs.
Best for: Classification, moderation, simple extraction, edge deployment, IoT
🟢 Mistral Small 4 — $0.15/$0.60
Mistral's workhorse budget model. Multimodal (text+image input), multilingual, Apache 2.0 licensed. Handles most production workloads that don't need the largest context windows. Strong at summarization, translation, and chat.
Best for: High-volume production, chatbots, summarization, translation, content moderation
🔵 Codestral — $0.30/$0.90
Purpose-built for code: completion, fill-in-the-middle, code generation, and debugging. 256K context window handles large codebases. Mistral proprietary license (not Apache 2.0).
Best for: Code completion, IDE integration, code review, automated testing
🔵 Mistral Large 3 — $0.50/$1.50
Mistral's open-weight flagship. 262K context (largest in the family), strong reasoning, multimodal, multilingual. Apache 2.0 licensed — self-host or use the API. Best balance of capability and cost for complex production workloads.
Best for: Complex reasoning, analysis, RAG, agentic workflows, enterprise production
🟣 Magistral Medium — $2.00/$5.00
Mistral's reasoning specialist. Designed for long-horizon tasks, synchronous tool-calling, and agentic coding. 256K context. Mistral proprietary license.
Best for: Multi-step reasoning, agentic coding, complex planning, research
Real-World Cost: 50K API Calls/Month
Estimated monthly cost for a production workload averaging 1,000 input tokens and 500 output tokens per request:
| Model | Input Cost | Output Cost | Total/Month | vs GPT-5.4 |
|---|---|---|---|---|
| Ministral 3 3B | $5.00 | $2.50 | $7.50 | 98% cheaper |
| Mistral Small 4 | $7.50 | $15.00 | $22.50 | 94% cheaper |
| Mistral Large 3 | $25.00 | $37.50 | $62.50 | 83% cheaper |
| Magistral Medium | $100.00 | $125.00 | $225.00 | 40% cheaper |
| GPT-5.4 (reference) | $125.00 | $375.00 | $375.00 | — |
EU Data Sovereignty: The Mistral Advantage
For organizations that must comply with GDPR or keep data within the EU, Mistral is the only major AI provider offering regional inference:
- EU data processing: All inference runs in EU data centers. +10% surcharge on standard pricing.
- No data transfer: Data never leaves the EU — no US Cloud Act exposure.
- GDPR compliant: Built for healthcare, finance, government, and regulated industries.
- Open-weight models: Self-host Mistral Large 3, Small 4, Ministral, and Devstral for full data control.
Compare Mistral Pricing Side-by-Side
Use our free calculator to compare Mistral models against 93 other AI models across 11 providers.
Open Calculator → Compare Models →Batch API & Caching
Mistral offers additional cost-saving options:
- Batch API: ~50% off standard pricing for async workloads. Submit batches of requests, results within 24 hours.
- Cached input tokens: 90% reduction on input costs for repeated context (RAG, system prompts).
- Enterprise pricing: 75% above list pricing on select APIs with SLAs and premium support.
Frequently Asked Questions
How much does the Mistral API cost in 2026?
Mistral has 11 models ranging from $0.10/$0.10 per 1M tokens (Ministral 3 3B) to $2.00/$5.00 (Magistral Medium). Budget options: Ministral 3 3B ($0.10/$0.10), Devstral Small 2 ($0.10/$0.30), Mistral Small 4 ($0.15/$0.60), Mistral Large 3 ($0.50/$1.50). All models support 128K–262K context.
Is Mistral cheaper than OpenAI?
Yes. Mistral Large 3 ($0.50/$1.50) is 80% cheaper than GPT-5.4 ($2.50/$15) for input and 90% cheaper for output. Mistral Small 4 ($0.15/$0.60) is 94% cheaper than GPT-5.4. Even Mistral's most expensive model (Magistral Medium at $2/$5) costs less than GPT-5.4.
Which Mistral model should I use?
For most production workloads: Mistral Large 3 ($0.50/$1.50, 262K context) — best balance of capability and cost. For high-volume/batch: Mistral Small 4 ($0.15/$0.60) or Ministral 3 3B ($0.10/$0.10). For code: Codestral ($0.30/$0.90, 256K). For reasoning: Magistral Medium ($2/$5, 256K). For edge/embedded: Ministral 3 3B.
Does Mistral offer EU data residency?
Yes. Mistral offers regional inference with EU data processing for a 10% surcharge. This makes Mistral the only major AI provider offering GDPR-compliant EU data residency for all models. Data stays in EU data centers — critical for healthcare, finance, and government workloads.
Are Mistral models open-weight?
Yes, most Mistral models are open-weight under Apache 2.0 license: Mistral Large 3, Mistral Small 4, Ministral 3 (3B/8B/14B), Devstral 2, and Devstral Small 2. You can self-host these models for zero API costs. Magistral and Codestral use Mistral's proprietary license.
What's the cheapest Mistral model?
Ministral 3 3B at $0.10/$0.10 per 1M tokens. It has symmetric pricing (input = output cost), making it especially cheap for output-heavy tasks like content generation. 128K context, Apache 2.0 licensed — self-host for free.